The tax intelligence layer behind SignalsHQ, now via MCP
For tax teams with or without developers. CPAs connect. Developers build. AI agents call.
INTEGRATION
Two ways in.
MCP servers
Drop directly into Claude, GPT-based coding agents, or an internal copilot your team is already building. Most teams are calling a capability within minutes of getting access.
REST API
For systems that aren’t running an MCP-compatible client yet. Same capabilities, plain HTTP.
FOUR MCP SERVERS
What you can call
Tax Engine
tax_engine.compute()
Send client return data in. Get a computed return back, with gaps and issues flagged before they become a problem in review.
Firms using the underlying engine see 70% less time spent in prep.
Document Intelligence
document.run()
Feed it the documents that break most OCR: P&L statements, K-1s with K-3s and footnotes, consolidated 1099s, rental income schedules. Get structured, usable data out.
98%+ extraction accuracy on complex tax documents.
Advisory Intelligence
advisory.run()
Send client data in. Get back advisory opportunities, but only from a list your firm controls and approves. Not open-ended AI advice. Nothing surfaces that your firm hasn’t already signed off on as a valid opportunity type.
Coming Soon
Intake & Review Intelligence
intake.run()
Client document and data intake as its own callable layer, the front door to everything above it.
Planned sources: Lacerte, ProConnect, UltraTax, CCH Axcess, Drake.
PATTERNS
What firms build with it
Compose with your own tools
Chain a SignalsHQ call into the systems your team already runs. The MCP returns structured data, so the next step is yours.
›
Extract the K-1s from this batch, then push the results into our internal review queue.
Build a return from client data
Feed in what the client sent, in whatever form it arrived. The engine works up from source documents to schedules to the 1040.
›
Here’s the client’s PBC folder and a JSON dump from our portal. Build the 1040 with every schedule it needs.
Check a return before review
Catch movement that usually only surfaces when a reviewer reads the whole file side by side.
›
Compare this return to last year and flag anything that moved more than 20%.
Sweep a book of clients
Ask across a whole segment instead of opening returns one at a time.
›
Across my Q1 returns, list every client still missing a 1099.
Interrogate a document
Footnotes and K-3 detail are where multi-state issues hide. Ask the document directly.
›
What’s in the footnotes of this K-1 that affects the state return?
Surface advisory opportunities
Answers come only from the opportunity list your firm has already approved.
›
Which of my clients hit a QBI threshold this year?
Secure Conversations. Smarter Tax Strategies.
Built on industry-leading standards to keep your clients' data safe and your firm audit-ready.
End-to-end Encryption
Monitoring Systems
Quarterly scans & annual pen tests
Access Control
Role‑based permissions